Everything AMD Revealed at Its Computex 2025 Press Conference in 19 Minutes
At Computex 2025 in Taipei, AMD delivered its most aggressive pitch yet to stake a claim as a top-tier AI innovator, spanning everything from budget GPUs to AI workstations and cloud-scale solutions.
This wasn’t just a hardware refresh. AMD’s presentation signaled a strategic transformation: it aims to dominate every layer of the AI stack—from consumer PCs and gaming systems to enterprise data centers and AI-powered edge devices.
Radeon RX 9060 XT: Affordable AI Gaming Gets Real
The show opened with a surprise launch: the Radeon RX 9060 XT, arriving June 5, priced at just $299 (8GB) and $349 (16GB). Billed as “the fastest GPU under $350,” the 9060 XT delivers competitive 1440p gaming and 821 TOPS of AI compute, underscoring AMD’s new mantra: AI-first, even in gaming.
Backed by next-gen FSR Redstone, AMD’s ML-enhanced upscaling engine, the 9060 XT offers features like neural radiance caching, ML super-resolution, and ray regeneration. It aims squarely at Nvidia’s DLSS with broader GPU compatibility.
Two-Pronged AI Strategy: Cloud to Client
AMD’s keynote hammered home a core message: the AI future is hybrid, not confined to the cloud.
-
AMD continues to scale in the cloud with its EPYC CPUs and Instinct GPUs powering hyperscalers like Meta and Netflix.
-
The star was the new Radeon AI PRO R9700 at the edge: a workstation GPU with 32GB of VRAM and PCIe Gen 5 support. With 4x the AI throughput of its predecessor, the R9700 is built for on-prem AI—local LLM fine-tuning, real-time video generation, and latency-sensitive tasks.
This push emphasizes AMD’s belief that edge AI will become foundational to the modern developer workflow—and that not all AI needs to live in a datacenter.
Ryzen AI 300 Series: Striking at the Heart of the AI PC Boom
On the client side, AMD unveiled the Ryzen AI 300 Series, including the Ryzen AI Max, which it claims outperforms Apple’s M4 Pro by 15%. With up to 50 TOPS of NPU performance and dedicated silicon for vision, audio, and language tasks, AMD is ferociously targeting the emerging AI PC market.
This launch aligns with Microsoft’s Copilot+ initiative and the broader industry shift toward on-device intelligence.
A Thriving Ecosystem Backed by Strategic Partners
AMD emphasized it isn’t moving alone—its AI ecosystem is expanding through critical alliances:
-
Lenovo is embedding Ryzen AI chips across its ThinkPad, ThinkBook, and Yoga lines.
-
Asus launched four Expert P Series business PCs using Ryzen AI PRO 300 chips with enterprise-grade security.
-
TSMC, AMD’s closest silicon partner, publicly praised the AI performance and battery life of AMD-powered laptops deployed internally.
These aren’t courtesy shout-outs—they’re a sign that AMD’s AI-first strategy is gaining serious traction across the supply chain.
Threadripper PRO 9000: The AI Workstation Renaissance
For professionals, AMD announced the Threadripper PRO 9000 WX-Series, led by the flagship 9995WX with 96 cores and 192 threads on Zen 5 architecture. It’s aimed directly at creative studios, engineers, and AI developers needing massive parallel compute.
AMD claims up to 80% better performance over rivals. To make the point, Weta FX—the VFX house behind Avatar and The Lord of the Rings—is switching its render farm to AMD-based systems, citing 60% performance gains.
AMD wants this platform—Threadripper plus Radeon AI PRO—to be the new baseline for local AI workstation deployments.
ROCm Comes to Windows: Bridging the Software Divide
In a pivotal move, AMD announced that ROCm, its AI software stack, is coming to Windows in the second half of 2025, with native support for PyTorch and ONNX.
This is critical: developers have long complained that AMD’s AI hardware lacked equivalent tooling to Nvidia’s CUDA. Bringing ROCm to Windows could open the floodgates for model training, inference, and tuning on AMD-powered desktops and laptops.
Headwinds and Hurdles Ahead
Despite the momentum, AMD still faces substantial challenges:
-
Nvidia’s developer ecosystem, CUDA dominance, and pretrained model support remain miles ahead.
-
Intel’s upcoming Lunar Lake chips, tightly integrated with Microsoft’s Copilot+ experience, could challenge AMD’s edge in AI PCs.
-
ROCm’s maturity and adoption will be the real litmus test. Open-source is a good start—ease of use and depth of support will determine success.
AMD’s AI Ambition: A Transformation in Progress
AMD’s Computex 2025 presentation wasn’t just about specs—it was a statement of intent. The company has moved from challenger to contender in AI, delivering a broad, coherent platform that spans client, cloud, and edge compute.
What remains to be seen is whether AMD can fully close the developer tools and ecosystem gap fast enough to capitalize on the AI boom.
Still, Lisa Su’s vision is coming into focus: AMD is no longer chasing the AI wave—it’s now building its own.
